[XPM] [ANN] Primecoin High Performance | HP14 released!

[Eli0t]

15

Are you using a service provider such Amazon? If yes, how much does it cost you to run that per hour?

4.5$ per hour

You inspired me to calculate a bit: If you get block at 10 mins average, block is 14.74 XPM now, you get 14.74 * 6 = 88.44 XPM/hour. if you pay 4.5$/hour it means you get 19.65 XPM for 1 US$!!! Wow! No wonder XPM exchange rate is 0.003 for BTC, you are getting 0.05896 BTC/US$, or 16.96US$ for 1BTC through XPM mining!!! Sadly but true, value of XPM should be more than 5x less, to be precise if BTC value is $95.40 now on MtGox, value of XPM should be 0.00053333 BTC. Shocking. Running to sell my poor XPM before people figure this out.

this is all based on "If you get block at 10 mins average," which i find very hard to belive

[1714503095]

1714503095

[1714503095]

1714503095

[1714503095]

1714503095

[crendore]

I might be missing a dependency. I'll take a look. I'm also using the following for CLFAGS:
CFLAGS = -mmacosx-version-min=10.7 -O3 -mfpmath=sse -march=native -mx32
Edit: it just occurred to me that the mx32 might be the issue. I will try with just native and o3 later.

Just tried compiling it with those exact flags:

Code:

CFLAGS = -mmacosx-version-min=10.7 -O3 -mfpmath=sse -march=native -mx32

And it worked fine. Only bug i have noticed so far is that it crashes when you do: primecoind stop
But that is a pretty minor problem.

-Crendore

[blastbob]

15

Are you using a service provider such Amazon? If yes, how much does it cost you to run that per hour?

4.5$ per hour

You inspired me to calculate a bit: If you get block at 10 mins average, block is 14.74 XPM now, you get 14.74 * 6 = 88.44 XPM/hour. if you pay 4.5$/hour it means you get 19.65 XPM for 1 US$!!! Wow! No wonder XPM exchange rate is 0.003 for BTC, you are getting 0.05896 BTC/US$, or 16.96US$ for 1BTC through XPM mining!!! Sadly but true, value of XPM should be more than 5x less, to be precise if BTC value is $95.40 now on MtGox, value of XPM should be 0.00053333 BTC. Shocking. Running to sell my poor XPM before people figure this out.

wow aren't you a smart one.

He just want cheap coins

[itod]

15

Are you using a service provider such Amazon? If yes, how much does it cost you to run that per hour?

4.5$ per hour

You inspired me to calculate a bit: If you get block at 10 mins average, block is 14.74 XPM now, you get 14.74 * 6 = 88.44 XPM/hour. if you pay 4.5$/hour it means you get 19.65 XPM for 1 US$!!! Wow! No wonder XPM exchange rate is 0.003 for BTC, you are getting 0.05896 BTC/US$, or 16.96US$ for 1BTC through XPM mining!!! Sadly but true, value of XPM should be more than 5x less, to be precise if BTC value is $95.40 now on MtGox, value of XPM should be 0.00053333 BTC. Shocking. Running to sell my poor XPM before people figure this out.

this is all based on "If you get block at 10 mins average," which i find very hard to belive

Guy has 200.000 pps, it's not hard to belive he is getting block in 10 mins average, I got two blocks today with my miserable 1500 pps.

[dudeguy]

Everybody already figured that out.

I haven't! Why didn't you tell me, you dirty bastards   , I had to figure it out myself.

So there is pretty much zero incentive to hold long term? It's just like every other ALT after LTC? Tank fest?

[geeknik]

I'm only running about 5K pps total split between 3 local machines and since about 8 hours ago when I installed the HP2 clients, I've already mined 28.98 XPM. Could be worse, because it's been 4 days of non-stop mining with the original client since I received anything.

[itod]

Everybody already figured that out.

I haven't! Why didn't you tell me, you dirty bastards   , I had to figure it out myself.

So there is pretty much zero incentive to hold long term? It's just like every other ALT after LTC? Tank fest?

No, we just to have to hit new equilibrium somewhere at the midpoint, more people are going to rent AWS resources at US$4.5 * 24 * 30 = US$3240 a month, and it will get few times more difficult to hit a block. Price will probably stabilize around 0.001 XPM/BTC. Since I don't have that kind of money, I'll stick to my 1500 pps and cross my fingers.

Edit:
If there is rich investor around here, I'll be happy to set things up for him and share a profit 

[dudeguy]

Everybody already figured that out.

I haven't! Why didn't you tell me, you dirty bastards   , I had to figure it out myself.

So there is pretty much zero incentive to hold long term? It's just like every other ALT after LTC? Tank fest?

No, we just to have to hit new equilibrium somewhere at the midpoint, more people are going to rent AWS resources at US$4.5 * 24 * 30 = US$3240 a month, and it will get few times more difficult to hit a block. Price will probably stabilize around 0.001 XPM/BTC. Since I don't have that kind of money, I'll stick to my 1500 pps and cross my fingers.

Edit:
If there is rich investor around here, I'll be happy to set things up for him and share a profit 

Why are people missing the obvious strategy. We should have sold (if you didn't already) at 0.01-0.007 and held until it bottomed out, buy back and watch it rise when scarcity becomes more prevalent. Everyone should have at least dumped initially and then bought back at a lesser price.

[mikaelh]

New version is out. It mainly adds a tuning parameter called -sievesize. Use it like this: primecoind -sievesize=2000000

I did some experiments with the sieve size and I was able to get significantly higher more blocks with a sieve size of 2M on a Core i7-4770k. There did seem to be some variance though.

The parameter will also affect your primespersec but it's always better to measure the block rate for now.

[itod]

Everybody already figured that out.

I haven't! Why didn't you tell me, you dirty bastards   , I had to figure it out myself.

So there is pretty much zero incentive to hold long term? It's just like every other ALT after LTC? Tank fest?

No, we just to have to hit new equilibrium somewhere at the midpoint, more people are going to rent AWS resources at US$4.5 * 24 * 30 = US$3240 a month, and it will get few times more difficult to hit a block. Price will probably stabilize around 0.001 XPM/BTC. Since I don't have that kind of money, I'll stick to my 1500 pps and cross my fingers.

Edit:
If there is rich investor around here, I'll be happy to set things up for him and share a profit 

Why are people missing the obvious strategy. We should have sold (if you didn't already) at 0.01-0.007 and held until it bottomed out, buy back and watch it rise when scarcity becomes more prevalent. Everyone should have at least dumped initially and then bought back at a lesser price.

Who said they missed it? Everybody is selling their XPM for 0.003 BTC like crazy, It's just us stupid who needed so much time to find the explanation.

The big winner here is also Amazon, they will sell quite a lot server time this month. At least until GPU miner version comes out. I take back everything I said about GPU moving the small guy out of mining. It's obvious that if GPU doesn't do it, cloud services will.

[reb0rn21]

@mikaelh
I presume you can`t do AVX2 optimization, or primecoin can not benefit from it?
pooler did AVX2 optimization on cpuminer and gain is ~90% with haswell

[ryen123]

New version is out. It mainly adds a tuning parameter called -sievesize. Use it like this: primecoind -sievesize=2000000

I did some experiments with the sieve size and I was able to get significantly higher more blocks with a sieve size of 2M on a Core i7-4770k. There did seem to be some variance though.

The parameter will also affect your primespersec but it's always better to measure the block rate for now.

Can the sievesize tuning parameter be added/tweaked in .conf file (for windows)?

[eule]

Mikael, how do you decide which sievesize to set for which cpu? Or just experimenting in testnet? How many testnet blocks are there btw, wanted to test too, but downloading those 20k+ blocks from only 3 connections takes ages...
@ryen: sievesize=1000000 should work.

[Vannicke]

Does the sievesize trigger work with "primecoin-qt.exe -sievesize=###"?

I tried and it didn't throw up, but that doesn't mean it worked either.

[SC123]

New version is out. It mainly adds a tuning parameter called -sievesize. Use it like this: primecoind -sievesize=2000000

I did some experiments with the sieve size and I was able to get significantly higher more blocks with a sieve size of 2M on a Core i7-4770k. There did seem to be some variance though.

The parameter will also affect your primespersec but it's always better to measure the block rate for now.

"Use it like this: primecoind -sievesize=2000000" <--- where exactly do I use this? Tried to type it in the console and got an error message.

[eule]

"Use it like this: primecoind -sievesize=2000000" <--- where exactly do I use this? Tried to type it in the console and got an error message.

You need to stop primecoind first.

[ryen123]

Looks like it works adding into .conf file. Drop in pps rate after increase sievesize. Thanks eule.

** Update
Definitely works. Tried increase to max sievesize 10M and my pps dropped from 3500+ to 60.. haha

[mikaelh]

@mikaelh
I presume you can`t do AVX2 optimization, or primecoin can not benefit from it?
pooler did AVX2 optimization on cpuminer and gain is ~90% with haswell

This is about primes, not about hashes.

[mikaelh]

Mikael, how do you decide which sievesize to set for which cpu? Or just experimenting in testnet? How many testnet blocks are there btw, wanted to test too, but downloading those blocks takes ages...
@ryen: sievesize=1000000 should work.

I'm guessing the optimal value may have something to do with L2 and L3 cache sizes.

Large sieves also take long to build which means a new block may arrive before it's complete.

[Vannicke]

@mikaelh
I presume you can`t do AVX2 optimization, or primecoin can not benefit from it?
pooler did AVX2 optimization on cpuminer and gain is ~90% with haswell

This is about primes, not about hashes.

On my A8 my PPS actually rose rather then fell with sievesize=2000000 in the conf, though it rose very minimally, which makes me wonder if it does anything at all.